ARM: Spectre-BHB workaround

Workaround the Spectre BHB issues for Cortex-A15, Cortex-A57,
Cortex-A72, Cortex-A73 and Cortex-A75. We also include Brahma B15 as
well to be safe, which is affected by Spectre V2 in the same ways as
Cortex-A15.

Reviewed-by: Catalin Marinas <catalin.marinas@arm.com>
Signed-off-by: Russell King (Oracle) <rmk+kernel@armlinux.org.uk>

arch/arm/include/asm/assembler.h

@@ -107,6 +107,16 @@
 	.endm
 #endif
 
+#if __LINUX_ARM_ARCH__ < 7
+	.macro	dsb, args
+	mcr	p15, 0, r0, c7, c10, 4
+	.endm
+
+	.macro	isb, args
+	mcr	p15, 0, r0, c7, r5, 4
+	.endm
+#endif
+
 	.macro asm_trace_hardirqs_off, save=1
 #if defined(CONFIG_TRACE_IRQFLAGS)
 	.if \save

arch/arm/include/asm/spectre.h

@@ -14,15 +14,19 @@ enum {
 	__SPECTRE_V2_METHOD_ICIALLU,
 	__SPECTRE_V2_METHOD_SMC,
 	__SPECTRE_V2_METHOD_HVC,
+	__SPECTRE_V2_METHOD_LOOP8,
 };
 
 enum {
 	SPECTRE_V2_METHOD_BPIALL = BIT(__SPECTRE_V2_METHOD_BPIALL),
 	SPECTRE_V2_METHOD_ICIALLU = BIT(__SPECTRE_V2_METHOD_ICIALLU),
 	SPECTRE_V2_METHOD_SMC = BIT(__SPECTRE_V2_METHOD_SMC),
 	SPECTRE_V2_METHOD_HVC = BIT(__SPECTRE_V2_METHOD_HVC),
+	SPECTRE_V2_METHOD_LOOP8 = BIT(__SPECTRE_V2_METHOD_LOOP8),
 };
 
 void spectre_v2_update_state(unsigned int state, unsigned int methods);
 
+int spectre_bhb_update_vectors(unsigned int method);
+
 #endif

arch/arm/include/asm/vmlinux.lds.h

@@ -116,11 +116,23 @@
  */
 #define ARM_VECTORS							\
 	__vectors_lma = .;						\
-	.vectors 0xffff0000 : AT(__vectors_start) {			\
-		*(.vectors)						\
+	OVERLAY 0xffff0000 : NOCROSSREFS AT(__vectors_lma) {		\
+		.vectors {						\
+			*(.vectors)					\
+		}							\
+		.vectors.bhb.loop8 {					\
+			*(.vectors.bhb.loop8)				\
+		}							\
+		.vectors.bhb.bpiall {					\
+			*(.vectors.bhb.bpiall)				\
+		}							\
 	}								\
 	ARM_LMA(__vectors, .vectors);					\
-	. = __vectors_lma + SIZEOF(.vectors);				\
+	ARM_LMA(__vectors_bhb_loop8, .vectors.bhb.loop8);		\
+	ARM_LMA(__vectors_bhb_bpiall, .vectors.bhb.bpiall);		\
+	. = __vectors_lma + SIZEOF(.vectors) +				\
+		SIZEOF(.vectors.bhb.loop8) +				\
+		SIZEOF(.vectors.bhb.bpiall);				\
 									\
 	__stubs_lma = .;						\
 	.stubs ADDR(.vectors) + 0x1000 : AT(__stubs_lma) {		\

arch/arm/kernel/entry-armv.S

@@ -1002,12 +1002,11 @@ vector_\name:
 	sub	lr, lr, #\correction
 	.endif
 
-	@
-	@ Save r0, lr_<exception> (parent PC) and spsr_<exception>
-	@ (parent CPSR)
-	@
+	@ Save r0, lr_<exception> (parent PC)
 	stmia	sp, {r0, lr}		@ save r0, lr
-	mrs	lr, spsr
+
+	@ Save spsr_<exception> (parent CPSR)
+2:	mrs	lr, spsr
 	str	lr, [sp, #8]		@ save spsr
 
 	@
@@ -1028,6 +1027,44 @@ vector_\name:
 	movs	pc, lr			@ branch to handler in SVC mode
 ENDPROC(vector_\name)
 
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+	.subsection 1
+	.align 5
+vector_bhb_loop8_\name:
+	.if \correction
+	sub	lr, lr, #\correction
+	.endif
+
+	@ Save r0, lr_<exception> (parent PC)
+	stmia	sp, {r0, lr}
+
+	@ bhb workaround
+	mov	r0, #8
+1:	b	. + 4
+	subs	r0, r0, #1
+	bne	1b
+	dsb
+	isb
+	b	2b
+ENDPROC(vector_bhb_loop8_\name)
+
+vector_bhb_bpiall_\name:
+	.if \correction
+	sub	lr, lr, #\correction
+	.endif
+
+	@ Save r0, lr_<exception> (parent PC)
+	stmia	sp, {r0, lr}
+
+	@ bhb workaround
+	mcr	p15, 0, r0, c7, c5, 6	@ BPIALL
+	@ isb not needed due to "movs pc, lr" in the vector stub
+	@ which gives a "context synchronisation".
+	b	2b
+ENDPROC(vector_bhb_bpiall_\name)
+	.previous
+#endif
+
 	.align	2
 	@ handler addresses follow this label
 1:
@@ -1036,6 +1073,10 @@ ENDPROC(vector_\name)
 	.section .stubs, "ax", %progbits
 	@ This must be the first word
 	.word	vector_swi
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+	.word	vector_bhb_loop8_swi
+	.word	vector_bhb_bpiall_swi
+#endif
 
 vector_rst:
  ARM(	swi	SYS_ERROR0	)
@@ -1150,8 +1191,10 @@ vector_addrexcptn:
  * FIQ "NMI" handler
  *-----------------------------------------------------------------------------
  * Handle a FIQ using the SVC stack allowing FIQ act like NMI on x86
- * systems.
+ * systems. This must be the last vector stub, so lets place it in its own
+ * subsection.
  */
+	.subsection 2
 	vector_stub	fiq, FIQ_MODE, 4
 
 	.long	__fiq_usr			@  0  (USR_26 / USR_32)
@@ -1184,6 +1227,30 @@ vector_addrexcptn:
 	W(b)	vector_irq
 	W(b)	vector_fiq
 
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+	.section .vectors.bhb.loop8, "ax", %progbits
+.L__vectors_bhb_loop8_start:
+	W(b)	vector_rst
+	W(b)	vector_bhb_loop8_und
+	W(ldr)	pc, .L__vectors_bhb_loop8_start + 0x1004
+	W(b)	vector_bhb_loop8_pabt
+	W(b)	vector_bhb_loop8_dabt
+	W(b)	vector_addrexcptn
+	W(b)	vector_bhb_loop8_irq
+	W(b)	vector_bhb_loop8_fiq
+
+	.section .vectors.bhb.bpiall, "ax", %progbits
+.L__vectors_bhb_bpiall_start:
+	W(b)	vector_rst
+	W(b)	vector_bhb_bpiall_und
+	W(ldr)	pc, .L__vectors_bhb_bpiall_start + 0x1008
+	W(b)	vector_bhb_bpiall_pabt
+	W(b)	vector_bhb_bpiall_dabt
+	W(b)	vector_addrexcptn
+	W(b)	vector_bhb_bpiall_irq
+	W(b)	vector_bhb_bpiall_fiq
+#endif
+
 	.data
 	.align	2
 

arch/arm/kernel/entry-common.S

@@ -153,13 +153,37 @@ ENDPROC(ret_from_fork)
  *-----------------------------------------------------------------------------
  */
 
+	.align	5
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+ENTRY(vector_bhb_loop8_swi)
+	sub	sp, sp, #PT_REGS_SIZE
+	stmia	sp, {r0 - r12}
+	mov	r8, #8
+1:	b	2f
+2:	subs	r8, r8, #1
+	bne	1b
+	dsb
+	isb
+	b	3f
+ENDPROC(vector_bhb_loop8_swi)
+
+	.align	5
+ENTRY(vector_bhb_bpiall_swi)
+	sub	sp, sp, #PT_REGS_SIZE
+	stmia	sp, {r0 - r12}
+	mcr	p15, 0, r8, c7, c5, 6	@ BPIALL
+	isb
+	b	3f
+ENDPROC(vector_bhb_bpiall_swi)
+#endif
 	.align	5
 ENTRY(vector_swi)
 #ifdef CONFIG_CPU_V7M
 	v7m_exception_entry
 #else
 	sub	sp, sp, #PT_REGS_SIZE
 	stmia	sp, {r0 - r12}			@ Calling r0 - r12
+3:
  ARM(	add	r8, sp, #S_PC		)
  ARM(	stmdb	r8, {sp, lr}^		)	@ Calling sp, lr
  THUMB(	mov	r8, sp			)

arch/arm/kernel/spectre.c

@@ -45,6 +45,10 @@ ssize_t cpu_show_spectre_v2(struct device *dev, struct device_attribute *attr,
 		method = "Firmware call";
 		break;
 
+	case SPECTRE_V2_METHOD_LOOP8:
+		method = "History overwrite";
+		break;
+
 	default:
 		method = "Multiple mitigations";
 		break;

arch/arm/kernel/traps.c

@@ -30,6 +30,7 @@
 #include <linux/atomic.h>
 #include <asm/cacheflush.h>
 #include <asm/exception.h>
+#include <asm/spectre.h>
 #include <asm/unistd.h>
 #include <asm/traps.h>
 #include <asm/ptrace.h>
@@ -801,6 +802,43 @@ static void flush_vectors(void *vma, size_t offset, size_t size)
 	flush_icache_range(start, end);
 }
 
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+int spectre_bhb_update_vectors(unsigned int method)
+{
+	extern char __vectors_bhb_bpiall_start[], __vectors_bhb_bpiall_end[];
+	extern char __vectors_bhb_loop8_start[], __vectors_bhb_loop8_end[];
+	void *vec_start, *vec_end;
+
+	if (system_state >= SYSTEM_FREEING_INITMEM) {
+		pr_err("CPU%u: Spectre BHB workaround too late - system vulnerable\n",
+		       smp_processor_id());
+		return SPECTRE_VULNERABLE;
+	}
+
+	switch (method) {
+	case SPECTRE_V2_METHOD_LOOP8:
+		vec_start = __vectors_bhb_loop8_start;
+		vec_end = __vectors_bhb_loop8_end;
+		break;
+
+	case SPECTRE_V2_METHOD_BPIALL:
+		vec_start = __vectors_bhb_bpiall_start;
+		vec_end = __vectors_bhb_bpiall_end;
+		break;
+
+	default:
+		pr_err("CPU%u: unknown Spectre BHB state %d\n",
+		       smp_processor_id(), method);
+		return SPECTRE_VULNERABLE;
+	}
+
+	copy_from_lma(vectors_page, vec_start, vec_end);
+	flush_vectors(vectors_page, 0, vec_end - vec_start);
+
+	return SPECTRE_MITIGATED;
+}
+#endif
+
 void __init early_trap_init(void *vectors_base)
 {
 	extern char __stubs_start[], __stubs_end[];

arch/arm/mm/Kconfig

@@ -851,6 +851,16 @@ config HARDEN_BRANCH_PREDICTOR
 
 	   If unsure, say Y.
 
+config HARDEN_BRANCH_HISTORY
+	bool "Harden Spectre style attacks against branch history" if EXPERT
+	depends on CPU_SPECTRE
+	default y
+	help
+	  Speculation attacks against some high-performance processors can
+	  make use of branch history to influence future speculation. When
+	  taking an exception, a sequence of branches overwrites the branch
+	  history, or branch history is invalidated.
+
 config TLS_REG_EMUL
 	bool
 	select NEED_KUSER_HELPERS

arch/arm/mm/proc-v7-bugs.c

@@ -177,6 +177,81 @@ static void cpu_v7_spectre_v2_init(void)
 	spectre_v2_update_state(state, method);
 }
 
+#ifdef CONFIG_HARDEN_BRANCH_HISTORY
+static int spectre_bhb_method;
+
+static const char *spectre_bhb_method_name(int method)
+{
+	switch (method) {
+	case SPECTRE_V2_METHOD_LOOP8:
+		return "loop";
+
+	case SPECTRE_V2_METHOD_BPIALL:
+		return "BPIALL";
+
+	default:
+		return "unknown";
+	}
+}
+
+static int spectre_bhb_install_workaround(int method)
+{
+	if (spectre_bhb_method != method) {
+		if (spectre_bhb_method) {
+			pr_err("CPU%u: Spectre BHB: method disagreement, system vulnerable\n",
+			       smp_processor_id());
+
+			return SPECTRE_VULNERABLE;
+		}
+
+		if (spectre_bhb_update_vectors(method) == SPECTRE_VULNERABLE)
+			return SPECTRE_VULNERABLE;
+
+		spectre_bhb_method = method;
+	}
+
+	pr_info("CPU%u: Spectre BHB: using %s workaround\n",
+		smp_processor_id(), spectre_bhb_method_name(method));
+
+	return SPECTRE_MITIGATED;
+}
+#else
+static int spectre_bhb_install_workaround(int method)
+{
+	return SPECTRE_VULNERABLE;
+}
+#endif
+
+static void cpu_v7_spectre_bhb_init(void)
+{
+	unsigned int state, method = 0;
+
+	switch (read_cpuid_part()) {
+	case ARM_CPU_PART_CORTEX_A15:
+	case ARM_CPU_PART_BRAHMA_B15:
+	case ARM_CPU_PART_CORTEX_A57:
+	case ARM_CPU_PART_CORTEX_A72:
+		state = SPECTRE_MITIGATED;
+		method = SPECTRE_V2_METHOD_LOOP8;
+		break;
+
+	case ARM_CPU_PART_CORTEX_A73:
+	case ARM_CPU_PART_CORTEX_A75:
+		state = SPECTRE_MITIGATED;
+		method = SPECTRE_V2_METHOD_BPIALL;
+		break;
+
+	default:
+		state = SPECTRE_UNAFFECTED;
+		break;
+	}
+
+	if (state == SPECTRE_MITIGATED)
+		state = spectre_bhb_install_workaround(method);
+
+	spectre_v2_update_state(state, method);
+}
+
 static __maybe_unused bool cpu_v7_check_auxcr_set(bool *warned,
 						  u32 mask, const char *msg)
 {
@@ -217,4 +292,5 @@ void cpu_v7_ca15_ibe(void)
 void cpu_v7_bugs_init(void)
 {
 	cpu_v7_spectre_v2_init();
+	cpu_v7_spectre_bhb_init();
 }